Research map / Detection and segmentation / Open-set and few-shot detection
Conditional Latent Diffusion Models for Zero-Shot Instance Segmentation
Maximilian Ulmer, Wout Boerdijk, Rudolph Triebel, Maximilian Durner
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the refine model after the coarse model (ablation) | OC-DiT · BOP TUDL | AP: 32.5 -> 59.4 | Improved |
| add the refine model after the coarse model (ablation) | OC-DiT · BOP LMO | AP: 29.6 -> 40.1 | Improved |
| add the refine model after the coarse model (ablation) | OC-DiT · BOP YCBV | AP: 68.6 -> 71.7 | Improved |
| train with random-interval positional encodings on 8 objects and test on 16, instead of training on 16 objects directly (ablation) | OC-DiT · YCBV | AP: 60.3 -> 65.2 | Improved |
| interpolate positional encodings at test time instead of random interval training (ablation) | OC-DiT · YCBV | AP: 65.2 -> 55.6 | Hurt |
| add Objaverse-based synthetic training data to MegaPose GSO (ablation) | OC-DiT · YCBV | AP: 56.3 -> 63.1 | Improved |
| train on all three datasets rather than any subset (ablation) | OC-DiT · YCBV | AP: 62.0 -> 65.5 | Improved |
| add GSO segmentation data to Objaverse data (ablation) | OC-DiT · YCBV | AP: 62.0 -> 62.0 | No clear effect |
| increase number of diffusion steps beyond 9 (ablation) | OC-DiT · YCBV | AP | No clear effect |
| vary noise discretization rho (ablation) | OC-DiT · YCBV | AP | No clear effect |
| increase test-time ensemble size with spatial augmentations (ablation) | OC-DiT · YCBV | AP | Improved |
| condition refine model on all objects (without true label) instead of the true object (ablation) | OC-DiT refine · TUDL | AP: 65.4 -> 56.8 | Hurt |
Closest papers by meaning
- Prompting Diffusion Models for Zero-Shot Instance Segmentation (NeurIPS 2026 · similarity 0.89)
- Unleashing the Potential of the Diffusion Model in Few-shot Semantic Segmentation (NeurIPS 2024 · similarity 0.89)
- DiffCut: Catalyzing Zero-Shot Semantic Segmentation with Diffusion Features and Recursive Normalized Cut (NeurIPS 2024 · similarity 0.88)
- High-Precision Dichotomous Image Segmentation via Probing Diffusion Capacity (ICLR 2025 · similarity 0.88)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.88)
- Explore In-Context Segmentation via Latent Diffusion Models (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
